What is Digg?Sponsored by Travelzoo
Take Advantage of Ridiculously Low Holiday Airfares view!
travelzoo.com - Flights $52 and up for Thanksgiving, Christmas & New Year. But move on it now.
118 Comments
- unixer, on 10/12/2007, -10/+41@newdigger
Your Name says it all... - gib0r, on 10/12/2007, -0/+23Because a Zettabyte is a lot of Pr0n.
- Recursiveness, on 10/12/2007, -1/+24"(But no, the entire world is so shortsighted as to believe that whatever the NEW limit is will last FOREVER! See: IPv6 versus IPv4)"
*sigh*
From:
http://en.wikipedia.org/wiki/Ipv6#128-bit_length
"IPv6 has enough room for 340,282,366,920,938,463,463,374,607,431,768,211,456 (340 undecillion, 282 decillion, 366 nonillion, 920 octillion, 938 septillion, 463 sextillion, 463 quintillion, 374 quadrillion, 607 trillion, 431 billion, 768 million, 211 thousand, 456) unique addresses. In exponential form, this equates to ~3.402823669*(10^38) unique addresses."
No, I'm afraid the sun will go out before we use up all those numbers. - GMorgan, on 10/12/2007, -3/+25ZFS was developed by Sun Microsystems, not Apple.
- TenebrousX, on 10/12/2007, -8/+27don't let the name fool you, he's been a member since September '05
- D3koy, on 10/12/2007, -3/+20I think the only real reason they haven't made a new filesystem is because they can't think of a cool acronym...
May I suggest LAZR low-accruing zetabyte resources...(I'm sure that acronym makes no sense, but you get to say laser, and I'd pay to have my computer run on lasers, or LAZR) - rinks, on 10/12/2007, -4/+20buried for childish "fag" label. come on man, you're old enough to work a computer, you should know better.
- carpespasm, on 10/12/2007, -0/+16i'm not sure you understand how much an internets worth of pr0n is...
- johnthedebs, on 10/12/2007, -4/+19This is probably the least attractive thing about ZFS. The fact that it doesn't have the same numerical limitations as 64-bit file systems is just a minor boon. The more interesting and exciting features on in the speed, ease of use, data integrity, etc:
-It uses a pooled storage model (Windows Home Server is said to use a similar model) so there are no volumes with arbitrary partitions set: Each file system has available (by default) the space of all the disks used in the pool.
-Files are always consistent on disk. The practical upshot is no disk check, ever. No need. ZFS' copy-on-write model allow the file systems to never become corrupt due to sudden power failure.
-Protection from data corruption since every single file on disk has a hash recorded along with it, which is checked whenever the file is read and updated if the file is changed. If for whatever reason the hash check fails, ZFS can automatically "heal" the file from a redundant copy.
-Live data scrubbing to verify every file on disk. This is the only kind of maintenance you'll ever have to do and guarantees that all your files are intact and as you expect them to be.
Etc, etc.
A quick Google search will reveal all this and a bunch more - and keep in mind that this is Sun's creation and is already available in Solaris 10. Great to see it gaining momentum and (supposedly) being supported by Apple's upcoming Leopard release. Exciting stuff, as far as file systems go. - ROFLance, on 10/12/2007, -7/+22Sooner or later? Try later. They've had 7 years to develop a new one, and they didn't.
- ROFLance, on 10/12/2007, -2/+16@Tony³
I never said it's not going to happen, just probably not anytime soon. - OmegaNine, on 10/12/2007, -8/+22OK so i am jealous of the mac fan boys now. I want this on Windows.
Well written atrial, worth the read by they way. - TenebrousX, on 10/12/2007, -11/+22" If we take a look back, virtual memory is the way that a computer manages the RAM that is installed into it. When an end-user install more RAM into the computer, he does not have to partition the RAM, nor tell the computer how much RAM he added. The whole virtual memory process does all these calculations, and the operating system and programs can easily request more RAM and get it without any effort."
this is not what virtual memory is... - TonyCubed, on 10/12/2007, -4/+15I think sooner or later, Microsoft will need to make a new filesystem. NTFS is alright, but there are better File Systems out there.
- GMorgan, on 10/12/2007, -0/+11Virtual memory is not necessarily involving a page file or swap partition though, when a process is given it's own memory segment that can be considered a form of virtual memory because it presents a process with a memory space that is abstract rather than actual. None of it need ever be on the HDD it's just that is the most common form of virtual memory. Another one is where the full address space is used despite there not being enough RAM to fill it, then the memory space wraps back around.
Anyway virtual memory is just a system by which an address space is presented by the system that doesn't necessarily match the hardware reality, this may or may not involve swap space. - ryland2, on 10/12/2007, -2/+12@ ROFLance
I dugg you up for the way you decided to spell his name... - inactive, on 10/12/2007, -0/+10A million terabytes. Basically enough for a couple hours of Ultra Porn.
- vondur, on 10/12/2007, -1/+11When will someone get around to porting it to Linux? (not me, I can barely program hello world on paper)
- raptordrew, on 10/12/2007, -3/+13Correct me if I'm wrong, but isn't this article incorrect on its claims about Virtual Memory? I thought virtual memory had to do with creating a paging file on a hard drive and being able to swap information between the RAM and paging file, not keep the user from having to "partition the RAM" or "tell the computer how much RAM he added."
I read wikipedia's article on the topic, and it sounds like I'm right, but if someone can affirm/correct me, thanks.
Also, berwiki, I agree with you... this article is crappy in trying to prove anything real. - pingwax, on 10/12/2007, -3/+12From wikipedia:
Virtual memory or virtual memory addressing is a memory management technique, used by computer operating systems, more common in multitasking OSes, wherein non-contiguous memory is presented to a software (aka process) as contiguous memory. This contiguous memory is referred to as the virtual address space.
If ZFS allows multiple drives in a system that appear to the software as a single contiguous disk space, that would seem to fit the definition above pretty well. - linuxeventually, on 10/12/2007, -5/+14They touted WinFS then killed it... linux enthusiasts would say ext2/3/4 or reiserserf but hey what do I know, right?
- carpespasm, on 10/12/2007, -3/+12no, microsoft started out saying that winfs was to be an entirely new beast and that longhorn would be coded from scratch, then they said winfs was built on ntfs, but might as well be a whole new beast, then it was claimed that it was still pretty much ntfs but with new neato stuff, then they said "well, it wasn't gonna be that different anyway" and axed it.
when they made their first claims though, the linux community got scared and started making open source software that would do what the MS marketing dept. said longhorn would do. Now we have reiserfs4 and the like. microsoft still has ntfs. - adinb, on 10/12/2007, -0/+7That would be a petabyte. 10^3 Petabytes would be an Exabyte. 10^3 Exabytes would be a Zettabyte. (side stepping KiB/MiB/GiB/TiB/PiB/EiB/ZiB 2^x notation)
- scotticus, on 10/12/2007, -1/+8Because lusting after girls hasn't worked out so far?
- Recursiveness, on 10/12/2007, -0/+6For the people burying my comment. Be aware that there's roughly 6 billion people in the world. If the earth's population tripled, or even quadrupled, and each person claimed 1 million IPv6 addresses, that'd just barely reach around half of the available ipv6 addresses out there.
- ozroy, on 10/12/2007, -0/+5The problem is ZFS is released under a license that is incompatible with the GPL. So it won't be ported to linux.
There is a project however to get it working under linux using FUSE meaning the filesystem will run in Userspace.
http://zfs-on-fuse.blogspot.com/ - BrainInAJar, on 10/12/2007, -1/+6You could always ditch linux & try Nexenta.... it's ubuntu with the solaris kernel, and it uses ZFS for /home by default
- TonyCubed, on 10/12/2007, -4/+9@linuxeventually
WinFS was an extension to NTFS. I'm talking about a completely different File System approach.
@Roflance
It will happen sooner or later. - Obvioustroll, on 10/12/2007, -2/+7berwiki,
I think your confusion is because programmers use the words "file system" to refer to two different things. On the one hand, a file system is an arrangement of sectors on a disk drive (or flash card, or CD or what ever) but on the other hand, a file system is also the software that manages those sectors. So, yes, it is "special software" that allows your computer to work with your IDE drives, just like FAT is, or ext3.
raptor,
You're right, as far as it goes, but VM systems do more than just swap memory out to disk. They also re-arrange memory in a dynamic way. Once upon a time you really did have to manually tell the computer how much RAM it had and where that RAM was. Many machines had special RAM that could only be used for special purposes, or RAM that was hidden unless you played programming tricks. - bobcrotch, on 10/12/2007, -1/+5As much as people hate microsoft, windows and NTFS, NTFS is a pretty resilient file system. How many times has your computer powered down unexpectedly? Rebooted? (yeah, windows lol) and how many times have you *NOT* had major problems with your disk? Maybe a chkdsk at startup? Tell me what happens if you pull that with ext2/3/4? Reiser is great because of the journaling but ext isn't nearly as forgiving as NTFS.
Now XFS, thats a real man's file system! - geronimo, on 10/12/2007, -0/+4With linux you have been able to use LVM for years and it allows for snapshotting. Just create an LVM volume, put ext3/reiser/etc on it, and you can snapshot entire LVM volumes. Plus you can use "storage pools" aka volumes and migrate data around. If you want to take out a disk in a pool aka volume, migrate all the data off the drive via an LVM command and take the disk out. Add space by putting a drive in, make the volume encompass the new disk then expand your filesystem. zfs combines the fs + volume manager so you simply add a disk, give it a command and it does everything for you.
You can also create a software raid mirror then put lvm on top of that then put reiser on top of that. But zfs does all this in one package.
One thing the author didn't mention is raid-z, which is a replacement for raid 5 that is actually slower but offers better data protection. But if you use a hardware RAID card, it's faster and just as reliable. Raid-z is geared towards those people who don't want to pay $300 for a RAID card.
The ZFS checksumming sounds useful. - maehem, on 10/12/2007, -0/+4One reason is ease of setup.
Just put a bunch of drives in the system,no formatting needed
and the raid setup took me a couple minutes and the
storage was immediatly available. I was done!
I previously built a Linux raid system and spent whole weekends trying to
figure out why it stopped working. My ZFS 6TB box of fun has had ZERO hiccups
and has survived four upgrades of the beta version of OpenSolaris 11. Yay ZFS! - cquinnd, on 10/12/2007, -0/+3The combination of Dynamic Disk support, with evolutionary improvments to NTFS (plus the potential to revive WinFS at some point); puts them in a position to work towards a better file system with what they already have... rather than trying to start from scratch.
I do agree that it will happen sooner or later, but it will likely come in the form of another installable file system, as opposed to a direct replacement for FAT32, or NTFS. - BrainInAJar, on 10/12/2007, -0/+3"I have had Sun raid cabinets blow things up using raid 5 WITH a hot spare"
Good thing ZFS supports double-parity RAID (raid6 aka raidz2) then now isn't it? - BrainInAJar, on 10/12/2007, -1/+4no... it's already a free filesystem. It's under CDDL ( similar to apache or xorg or mozilla )
here, have the source if you want:
http://src.opensolaris.org/source/xref/onnv/onnv-gate/usr/src/uts/common/fs/zfs/ - alej744, on 10/12/2007, -0/+3because of the Z, geeks love their Zs
- mike503, on 10/12/2007, -2/+5ZFS is the best thing ever. once they add in encryption and secure deletion, it will also solve everything a filesystem could possibly handle (short of being cluster-aware / multi-reader and/or writer like GFSes)
i have recently been reading up more on ZFS and am about to purchase some hardware to install opensolaris and take advantage of it. i can't wait until linux and freebsd (and who knows what else) have fully functional production versions of it as well.
they even include a simple web-based UI for some basic functions, if they beefed it up more, it would be even easier for people to adopt this.
like it says it is the last filesystem anyone will need, ever. i tend to agree, from everything that is written about it. i especially like the fact it requires no hardware RAID (although it can't hurt) or exotic configurations. as long as they keep up on the performance front (especially re: encryption, which i suppose could be helped by encryption add-on cards?) this is the best thing since sliced bread.
and if anyone doubts pooled storage, look at what the windows home server does - the same idea. pooled storage (although obviously not as robust, and has more bugs or "gotchas" than ZFS i'm sure) - otomo, on 10/12/2007, -1/+4Sorry, no it can't. And zfs isn't proprietary.
The current work on getting zfs into, gasp, FreeBSD.
http://lists.freebsd.org/pipermail/freebsd-current/2006-August/065306.html - BrainInAJar, on 10/12/2007, -0/+2License incompatible.
Try nexenta or Belenix. they've already got it (since they're OpenSolaris distros) - berwiki, on 10/12/2007, -1/+3Hey Obvious Troll EAT A DICK!!!
"ZFS is currently not available as a root filesystem since there is no ZFS boot support. The ZFS Boot project is currently working on adding root filesystem support.[6]" - Wikipedia
Looks like you were wrong about your "file system" analysis. I knew exactly what the problem was with my questions above, and you just threw jargon in my face. Because ZFS is not like EXT3 or NTFS. - SVPirate, on 10/12/2007, -1/+3I'm so glad Apple got a hold of this. HFS+ Royally sucks compared to most other modern FSs. The Journalling support and Metadata (spotlight) have been bolted to an already creaking and old File structure that was introduced in the Late 80's (as plain HFS) and only recieved a major overhaul in 1995(or was it 96?) (HFS+) to support larger volumes and smaller block sizes. I nearly left the room via a hole in the roof with joy when I read that ZFS really WAS in 10.5. I've played with it on a small disk array with Solaris 10 (11/06) a Sun Ultra 60 workstation and I can tell you it rocks bells, big ones, that go CLANG!! From Apple's POV it's a perfect companion to Time Machine becuase it supports Snapshots and Mirroring without all the hassle of LVM or RAID, and the ability to add storage at will.
I want to see ZFS everywhere, but Linux distros would be my next target, especially now Sun plans to GPL OpenSolaris. It smokes most exisiting FSs for speed, usability and flexibility. - GMorgan, on 10/12/2007, -1/+3WinFS was just about building a relational database onto the FS with an AI scanning through your folders and files metatagging everything with the whole thing used as a mini data warehouse/mining operation in support of desktop search. It would have been extremely cool but MS has seemingly put it on indefinite suspension. Nothing in Reiser4 does any of this as such (it is good at blowing up your files though ;) ) and though Linux does have some neat desktop search tools they aren't anywhere near what MS proposed as of yet unfortunately. It will get there though and likely before WinFS arrives in SP* or more likely Vista +1.
- thesixthdesign, on 10/12/2007, -0/+2This article makes absolutely no sense to me at all (or perhaps I was merely bored after all the intricate use of vocabulary that I discontinued reading), yet there are so many Diggs.
And it's not like I am never on the computer, I am on the computer almost everyday.
Quite strange to me ... - gregrich, on 10/12/2007, -1/+3Because we like the fact that we are limited by the 'boiling of the earths oceans'....
- tedc, on 10/12/2007, -0/+2That article made it sound like any drawbacks in switching to ZFS would easily be outweighed by the benefits, but I still have to wonder what drawbacks exist. When I first heard about ZFS, they said it was not suitable for boot drives and had no support for low-level encryption. Maybe that has changed since, but what about HFS+ features like multiple forks or extensible meta-data? Would they translate well over to the ZFS world? I have to admit, though, that it does seem like a very good fit for Time Machine on a large, external backup drive.
- doublebackslash, on 10/12/2007, -0/+2I tend to disagree with the idea that error correction should be brought into kernel space. Sure, knowing your data is okay is great and all, but that is what error correction on the drive is for. Not enough? SAS and SATA support error detection and reporting on the line. Not enough? At this point you have to distrust your memory, there is ECC for that. That leaves the processor, and that is what the kernel space error checking will use. Sorry, ECC on the FS level is overkill if the hardware is done right, and it is done right in almost every place it counts.
- bigtoe416, on 10/12/2007, -1/+3"ZFS has no theoretical limit for a single file, but the whole filesystem can be as big as one zettabyte (a zettabyte is one million terabytes)"
Uhm, so isn't the theoretical limit for a single file...one zettabyte? Exactly. - zoom1928, on 10/12/2007, -1/+3> No, I'm afraid the sun will go out before we use up all those numbers.
You're forgetting that they're being assigned in large, inefficient blocks. With IPv4, the addresses are allocated in blocks to reduce the number of routes as they'll have to be w/ v6. There are plenty of v4 addresses left. There are only a few IPv4 blocks left. From looking at the number of addresses assigned in the past year, there's only 3 years and ten months left at the current rate even though most of the addresses already assigned are not used. With IPv6 there's an additional problem. Each host has a minimum of over 4 billion times 4 billion addresses(64 bits.) Of the remaining 64, 16 are used for subnets. That gives you 64k networks, and it only leaves 48 bits to identify blocks of addresses. Your math is incorrect. The number you gave was 2^128 rather than 2^48. - Obvioustroll, on 10/12/2007, -0/+2That's what I thought, too, but according to the various apple sites "Time Machine" does not depend on ZFS.
- SuperSunny, on 10/12/2007, -0/+2Funny...mine is exactly the reverse. I have terribly slow copy times on all the XP machines I've used, and suddenly copying on my MacBook is 10x faster...
-
Show 51 - 100 of 118 discussions
© Digg Inc. 2009
— Content created and posted by Digg users is